[grepjava]

When a bridge-to-address policy uses durable subscriptions, AMQPBridgeToAddressSenderController.createServerConsumer builds the subscription queue in a local QueueConfiguration named configuration, which shadows the controller's AMQPBridgeSenderConfiguration. Its three auto-delete lines copied the new queue's own unset values onto itself, so auto-delete-durable-subscription, auto-delete-durable-subscription-message-count and auto-delete-durable-subscription-delay were never applied. The queue is created as auto-created, so the address settings filled the gaps: with the defaults, auto-delete-queues (true) removed the bridge's durable subscription once its link was gone and the queue was empty, and messages sent while the remote was unreachable were not kept for it.

The new queue's auto-delete values now come from the sender configuration (this.configuration), as AMQPFederationAddressSenderController sets them from its own.

AMQPBridgeToAddressTest.testBridgeCreatesDurableBindingWhenSenderConfiguredToUseDurableSubscriptions now runs with a 100 ms address queue scan period and checks the queue is not auto-delete and is still there after the peer goes. testBridgeToAddressCleansUpDurableBindingIfConfiguredTo now checks the queue has the configured delay and message count. Both fail without the fix: before it, the first queue was auto-deleted and the second passed only through the address settings' defaults.

https://issues.apache.org/jira/browse/ARTEMIS-6328

[tabish121]

This rename change is unnecessary and obscures the actual changes that were needed, please revert that part.

[grepjava]

Reverted the rename. The local is configuration again, and the three auto-delete lines now read the sender's settings through this.configuration, since the local shadows the field. The main change is now just those three lines.

[tabish121]

Change seems fine now, need to squash the commits into one commit

[grepjava]

The DescribeJournalTest failures are from main (they fail on every main build since ARTEMIS-6242). The MirroredVersionTest.testTopic failure in the first run is unrelated too. That test configures only mirroring, and the changed code runs only for an AMQP bridge-to policy.